KOMPARASI ALGORITMA MACHINE LEARNING UNTUK MEMPREDIKSI DEPRESI PADA MAHASISWA MENGGUNAKAN TEKNIK SMOTE
Abstract
Depression is one of the most common mental health issue experienced by university students and can negatively affect both academic performance and quality of life. This study aims to compare the performance of several machine learning algorithms, namely Random Forest, XGBoost, and Logistic Regression, in predicting depression among students, as well as to analyze the impact of applying the Shynthetic Monority Oversampling Technique (SMOTE) to address class imbalance. The dataset used in this study was the Student Depression Dataset obtained from Kaggle, consisting of 27,901 records and 13 features after filtering the Profession attribute to include only students. The research methodology involved data preprocessing, data splitting with an 80:20 ratio, the application of SMOTE to the training data, model development using the three machine learning algorithms, and model evaluation based on accuracy, precision, recall, and F1-score metrics. The results indicate that Logistic Regression achieved the best performance,with an accuracy of 85.68%, precision of 87.15%, recall of 88.60%, and an F1-score of 87.87%. The implementation of SMOTE technique successfully balanced the class distribution in the training data from 58.45% and 41.55% to 50% and 50%, and consistently improved the precision, recall, and F1-score values across all three algorithms.

